https://doi.org/10.5194/esd-2019-43 Preprint. Discussion started: 9 September 2019 c Author(s) 2019. CC BY 4.0 License. 1 Groundwater storage dynamics in the world’s large aquifer systems 2 from GRACE: uncertainty and role of extreme precipitation 3 Mohammad Shamsudduha 1,2,* and Richard G. Taylor 1 4 1 Department of Geography, University College London, London, UK 5 2 Department of Geography, University of Sussex, Falmer, Brighton, UK 6 * Corresponding author: M. Shamsudduha ([email protected] ) 7 8 Abstract 9 Under variable and changing climates groundwater storage sustains vital ecosystems and 10 enables freshwater withdrawals globally for agriculture, drinking-water, and industry. Here, 11 we assess recent changes in groundwater storage ( ΔGWS) from 2002 to 2016 in 37 of the 12 world’s large aquifer systems using an ensemble of datasets from the Gravity Recovery and 13 Climate Experiment (GRACE) and Land Surface Models (LSMs). Ensemble GRACE- 14 derived ΔGWS is well reconciled to in-situ observations ( r = 0.62–0.86, p value <0.001) for 15 two tropical basins with regional piezometric networks and contrasting climate regimes. 16 Trends in GRACE-derived ΔGWS are overwhelmingly non-linear; indeed linear declining 17 trends adequately ( R2 >0.5, p value <0.001) explain variability in only two aquifer systems. 18 Non-linearity in ΔGWS at the scale of GRACE (~200,000 km 2) derives, in part, from the 19 episodic nature of groundwater replenishment associated with extreme annual (>90 th 20 percentile, 1901–2016) precipitation and is inconsistent with prevailing narratives of global- 21 scale groundwater depletion. Substantial uncertainty remains in estimates of GRACE-derived 22 ΔGWS, evident from 20 realisations presented here, but these data provide a regional context 23 to changes in groundwater storage observed more locally through piezometry. 24 1 https://doi.org/10.5194/esd-2019-43 Preprint. Discussion started: 9 September 2019 c Author(s) 2019. CC BY 4.0 License. 25 1 Introduction 26 Groundwater is estimated to supply substantial proportions of the world’s agricultural (42%), 27 domestic (36%), and industrial (27%) freshwater demand (Döll et al., 2012). As the world’s 28 largest distributed store of freshwater, groundwater also plays a vital role in sustaining 29 ecosystems and enabling adaptation to increased variability in rainfall and river discharge 30 brought about by climate change (Taylor et al., 2013a). Sustained reductions in the volume of 31 groundwater (i.e. groundwater depletion) resulting from human withdrawals or changes in 32 climate have historically been observed as declining groundwater levels recorded in wells 33 (Scanlon et al., 2012a; Castellazzi et al., 2016; MacDonald et al., 2016). The limited 34 distribution and duration of piezometric records hinder, however, direct observation of 35 changes in groundwater storage globally including many of the world’s large aquifer systems 36 (WHYMAP and Margat, 2008). 37 Since 2002 the Gravity Recovery and Climate Experiment (GRACE) has enabled large-scale 38 (≥ 200,000 km 2) satellite monitoring of changes in total terrestrial water storage ( ΔTWS) 39 globally (Tapley et al., 2004). As the twin GRACE satellites circle the globe ~15 times a day 40 they measure the inter-satellite distance at a minute precision (within one micron) and 41 provide ΔTWS for the entire earth approximately every 30 days. GRACE satellites sense 42 movement of total terrestrial water mass derived from both natural (e.g. droughts) and 43 anthropogenic (e.g. irrigation) influences globally (Rodell et al., 2018). Changes in 44 groundwater storage (GRACE-derived ΔGWS) are computed from ΔTWS after deducting 45 contributions (equation 1) that arise from other terrestrial water stores including soil moisture 46 (ΔSMS), surface water ( ΔSWS), and the snow water storage (ΔSNS) using data from Land 47 Surface Models (LSMs) either exclusively (Rodell et al., 2009; Famiglietti et al., 2011; 48 Scanlon et al., 2012a; Famiglietti and Rodell, 2013; Richey et al., 2015; Thomas et al., 2017) 2 https://doi.org/10.5194/esd-2019-43 Preprint. Discussion started: 9 September 2019 c Author(s) 2019. CC BY 4.0 License. 49 or in combination with in situ observations (Rodell et al., 2007; Swenson et al., 2008; 50 Shamsudduha et al., 2012). 51 ΔGWS = ΔTWS – (ΔSMS + ΔSWS + ΔSNS) (1) 52 Substantial uncertainty persists in the quantification of changes in terrestrial water stores 53 from GRACE measurements that are limited in duration (2002 to 2016), and the application 54 of uncalibrated, global-scale LSMs (Shamsudduha et al., 2012; Döll et al., 2014; Scanlon et 55 al., 2018). Computation of ΔGWS from GRACE ΔTWS is argued, nevertheless, to provide 56 evaluations of large-scale changes in groundwater storage where regional-scale piezometric 57 networks do not currently exist (Famiglietti, 2014). 58 Previous assessments of changes in groundwater storage using GRACE in the world’s 37 59 large aquifer systems (Richey et al., 2015; Thomas et al., 2017) (Fig. 1, Table 1) have raised 60 concerns about the sustainability of human use of groundwater resources. One analysis 61 (Richey et al., 2015) employed a single GRACE ΔTWS product (CSR) in which changes in 62 subsurface storage ( ΔSMS + ΔGWS) were attributed to ΔGWS. This study applied linear 63 trends without regard to their significance to compute values of GRACE-derived ΔGWS over 64 11 years from 2003 to 2013, and concluded that the majority of the world’s aquifer systems 65 (n=21) are either “overstressed” or “variably stressed”. A subsequent analysis (Thomas et al., 66 2017) employed a different GRACE ΔTWS product (Mascons) and estimated ΔSWS from 67 LSM data for both surface and subsurface runoff, though the latter is normally considered to 68 be groundwater recharge (Rodell et al., 2004). Using performance metrics normally applied 69 to surface water systems including dams, this latter analysis classified nearly a third ( n=11) of 70 the world’s aquifer systems as having their lowest sustainability criterion. 71 Here, we update and extend the analysis of ΔGWS in the world’s 37 large aquifer systems 72 using an ensemble of three GRACE ΔTWS products (CSR, Mascons, GRGS) over a 14-year 3 https://doi.org/10.5194/esd-2019-43 Preprint. Discussion started: 9 September 2019 c Author(s) 2019. CC BY 4.0 License. 73 period from August 2002 to July 2016. To isolate GRACE-derived ΔGWS from GRACE 74 ΔTWS, we employ estimates of ΔSMS, ΔSWS and ΔSNS from five LSMs (CLM, Noah, 75 VIC, Mosaic, Noah v.2.1) run by NASA’s Global Land Data Assimilation System (GLDAS). 76 As such, we explicitly account for the contribution of ΔSWS to ΔTWS, which has been 77 commonly overlooked (Rodell et al., 2009; Richey et al., 2015; Bhanja et al., 2016) despite 78 evidence of its significant contribution to ΔTWS (Kim et al., 2009; Shamsudduha et al., 79 2012; Getirana et al., 2017). Further, we characterise trends in time-series records of 80 GRACE-derived ΔGWS by employing a non-parametric, Seasonal-Trend decomposition 81 procedure based on Loess (STL) (Cleveland et al., 1990) that allows for resolution of 82 seasonal, trend and irregular components of GRACE-derived ΔGWS for each large aquifer 83 system. In contrast to linear or multiple-linear regression-based techniques, STL assumes 84 neither that data are normally distributed nor that the underlying trend is linear 85 (Shamsudduha et al., 2009; Humphrey et al., 2016; Sun et al., 2017). 86 87 2 Data and Methods 88 2.1 Global large aquifer systems 89 We use the World-wide Hydrogeological Mapping and Assessment Programme (WHYMAP) 90 Geographic Information System (GIS) dataset for the delineation of world's 37 Large Aquifer 91 Systems (Fig. 1, Table1) (WHYMAP and Margat, 2008). The WHYMAP network, led by 92 the German Federal Institute for Geosciences and Natural Resources (BGR), serves as a 93 central repository and hub for global groundwater data, information, and mapping with a goal 94 of assisting regional, national, and international efforts toward sustainable groundwater 95 management (Richts et al., 2011). The largest aquifer system in this dataset (Supplementary 96 Table S1) is the East European Aquifer System (WHYMAP no. 33; area: 2.9 million km 2) 4 https://doi.org/10.5194/esd-2019-43 Preprint. Discussion started: 9 September 2019 c Author(s) 2019. CC BY 4.0 License. 97 and the smallest one the California Central Valley Aquifer System (WHYMAP no. 16; area: 98 71,430 km 2), which is smaller than the typical sensing area of GRACE (~200,000 km 2). 99 However, Longuevergne et al. (2013) argue that GRACE satellites are sensitive to total mass 100 changes at a basin scale so ΔTWS measurements can be applied to smaller basins if the 101 magnitude of temporal mass changes is substantial due to mass water withdrawals (e.g., 102 intensive groundwater-fed irrigation). Mean and median sizes of these large aquifers are 103 ~945,000 km 2 and ~600,000 km 2, respectively. 104 2.2 GRACE products 105 We use post-processed, gridded (1° × 1°) monthly GRACE TWS data from CSR land 106 (Landerer and Swenson, 2012) and JPL Global Mascon (Watkins et al., 2015; Wiese et al., 107 2016) solutions from NASA’s dissemination site (http://grace.jpl.nasa.gov/data), and a third 108 GRGS GRACE solution (CNES/GRGS release RL03-v1) (Biancale et al., 2006) from the 109 French Government space agency, Centre National D'études Spatiales (CNES). To address 110 the uncertainty associated with different GRACE processing strategies (CSR, JPL-Mascons, 111 GRGS), we apply an ensemble mean of the three GRACE solutions (Bonsor et al., 2018). 112 CSR land solution (version RL05.DSTvSCS1409) is post-processed from spherical 113 harmonics released by the Centre for Space Research (CSR) at the University of Texas at 114 Austin.